53. “The EU Policy Machine Is Broken. Brussels Lacks Courage!” - Kai Zenner on Europe’s Digital Future and the Steps We Must Take in 2026 to Reclaim It

1 - > Alexandra Ebert: Hello and welcome to the AI and Data 2 - > Democratization Podcast. 3 - > I'm your host, Alexandra Ebert, and with today's episode, I'm 4 - > super excited to kick us off into the sixth season of the 5 - > show. 6 - > Six seasons, can you believe that? 7 - > What awaits us over the next few months is, without question, 8 - > the best, boldest, and most brilliant guest lineup we've had 9 - > so far.

10 - > But no, I'm not revealing our upcoming guests just yet. 11 - > After all, where has there ever been any fun in spoilers? 12 - > What I am going to tell you, however, is that I have a big 13 - > confession to make. 14 - > Now, the most attentive listeners among you might have 15 - > already seen this coming.

16 - > What started as a flirt, a little mention here, a little 17 - > mention there, over time turned into something okay, who am I 18 - > kidding? 19 - > This was never something casual. 20 - > And we were pretty much head over heels right from the start. 21 - > And you know how these things go.

22 - > After a while, you ask yourself, should you continue 23 - > treating them as this unofficial plus one, or finally give them 24 - > the recognition that they deserve? 25 - > So after more than five years together, I figured it was only 26 - > fair to make things official. 27 - > What am I talking about here? 28 - > AI, of course.

29 - > Because let's be honest, the data democratization podcast has 30 - > always been about data and AI democratization and how we can 31 - > make both of them work for more people, more businesses, and 32 - > societies at large. 33 - > Around here, AI has never been the sidekick. 34 - > It was always core to the conversations I've had on this 35 - > show. 36 - > And I'm not planning to change that in the future.

37 - > So, starting with this first episode of our new season, I'm 38 - > officially promoting AI into the title of this show. 39 - > And from here on out, I'm happy to welcome you to my bi-weekly 40 - > episodes of the AI and data democratization podcast. 41 - > But enough about naming conventions and to the real 42 - > reason you are all actually here today. 43 - > Let's join me in welcoming Kai Zenner on the show, who is 44 - > Head of Office and Digital Policy Adviser to Axel Voss at 45 - > the European Parliament, and one of the best sources to chat 46 - > about the 2026 AI and data policy landscape in the European 47 - > Union.

48 - > And with everything that's on the agenda, 2026 promises to be 49 - > a pivotal year here for us in the European Union. 50 - > We recorded this conversation end of January this year and 51 - > talked about a broad range of topics, from Kai's thoughts on 52 - > Davos to his landmark report on The European Way and on 53 - > Reclaiming Europe's Digital Future, to his criticism on the 54 - > EU policy machine and why Europe needs to demonstrate courage 55 - > and bold political leadership to find its way in this rapidly 56 - > evolving global digital economy.

57 - > So without further ado, let us dive in. 58 - > I was so looking forward to our conversation, and I can't wait 59 - > to dive into all the exciting topics we have for today. 60 - > But before we really do that, as always, I ask my guests to 61 - > introduce themselves. 62 - > So can you let our listeners know who you are, even though 63 - > probably every one of them knows you already and what you're 64 - > doing as the Head of Office and Digital Policy Adviser to Axel 65 - > Voss from the European Parliament?

66 - > Kai Zenner: Yes, of course. 67 - > First of all, many thanks for the invitation. 68 - > And indeed, you said it. 69 - > I'm working in the European Parliament for MEP Axel Voss.

70 - > The two of us now for yeah um over eight years are working 71 - > together on AI and um data policy. 72 - > Um we were very, very strongly involved in the GDPR, but also 73 - > then in the European AI Act. 74 - > We also tried to push this AI Liability Directive or 75 - > regulation that unfortunately didn't make it. 76 - > Um but the two of us um also are now for a long time working 77 - > on tech sovereignty or tech resilience in Europe.

78 - > Um, myself, I um co-started the Eurostack movement um two years 79 - > ago. 80 - > And yeah, so a lot of uh topics, but in the end, it's all 81 - > going um back to how do we accelerate uh Europe's 82 - > digitization. 83 - > Alexandra Ebert: That's definitely something that can 84 - > keep one very, very busy. 85 - > Another standard question I have on the data democratization 86 - > podcast because I find it always interesting to figure out 87 - > why people are doing what they're doing.

88 - > So I have many ideas why you're in your role, but what drives 89 - > you to do this important work that you have been doing in the 90 - > past eight years. 91 - > Kai Zenner: Yeah. 92 - > At the beginning, there was definitely an element of chance. 93 - > Um, so I started my professional life at Konrad 94 - > Adenauer Foundation, and digital policy was just one of many 95 - > items that I uh covered.

96 - > But yeah, it um it was super interesting, it was very 97 - > international uh from the beginning, uh, this area. 98 - > Uh therefore I liked it. 99 - > And then Axe and I were basically the almost the 100 - > earliest contenders for AI policy in the European 101 - > Parliament because we liked it because we are two lawyers, uh 102 - > both of us are lawyers, and um we saw a lot of very interesting 103 - > uh legal questions uh that uh new AI systems are posing to our 104 - > democratic system, but also to the economy, and so on, and so 105 - > on.

106 - > And yeah, this is how we got into the topic, and from my 107 - > character, I'm always trying to dive super deep and learn 108 - > everything and so on. 109 - > So AI became over the years, AI policy became basically almost 110 - > a hobby for me. 111 - > I can imagine. 112 - > Yeah, I really try to push wherever I have the chance for.

113 - > Alexandra Ebert: Yeah, that's also definitely the impression 114 - > when we got to know each other five years ago. 115 - > I can't even remember when we started uh collaborating, also 116 - > talking about the AI Act. 117 - > That I really found it admirable that both you and Axel 118 - > Voss had this deep ambition and drive to bring in solid 119 - > perspective, technical voices, um, all the different 120 - > stakeholders that are needed to shape and determine Europe's 121 - > digital future.

122 - > Because, as we all know, AI is not just one new technological 123 - > innovation, it's really a foundational shift that is 124 - > impacting so many large layers of the societal fabric. 125 - > So really glad that the European Parliament has uh such 126 - > uh competent people as you and Axel Voss working on these 127 - > important topics. 128 - > Today I predominantly want to talk about the role of the 129 - > European Union and Europe in this entire AI and data and 130 - > digital ecosystem.

131 - > Since we're recording this in the last week of January and 132 - > you're just freshly returned from Davos, let's maybe start 133 - > off with your experience from there. 134 - > What are the headline thoughts and insights that you took away 135 - > from what you observed there and shared there? 136 - > Kai Zenner: Yeah. 137 - > Um it was now the third time um that I was in Davos, so um it 138 - > started basically uh three years ago.

139 - > Um I'm often there with the AI house, but also with some other 140 - > um side events, and um normally it's a very positive atmosphere, 141 - > um, yeah, similar as you know it from New York, United 142 - > Nations, or UNESCO, OECD in Paris, and so on. 143 - > So everyone talks with each other and so on. 144 - > This time, I think, especially the first two days, in 145 - > particular, the Europeans were kind of afraid what Donald Trump 146 - > will say, how the Americans will position themselves, and 147 - > therefore, yeah, there was a lot of anxiety, let's say.

148 - > After the speech from Trump, um, a bit of relief that certain 149 - > things didn't happen, that it was a bit more cooperative, but 150 - > you can definitely summarize it with there are high geopolitical 151 - > tensions, in particular between um those historical partners in 152 - > the EU and in the United States, and uh why Davos has 153 - > helped to calm down both sides a bit. 154 - > Um there is a lot of things that needs to be fixed that 155 - > needs to be rediscussed, and so on.

156 - > And um yeah, uh, Davos also showed that um different places 157 - > in the world have so different opinions how we should uh 158 - > proceed that this, what I just said, will become extremely 159 - > difficult. 160 - > Because yeah, there basically what is missing nowadays is this 161 - > kind of um joint spirit to um address certain global 162 - > challenges. 163 - > But yeah, we are kind of back to what I always say neorealism, 164 - > and so everyone is fighting for himself or herself.

165 - > Alexandra Ebert: Yeah, which is not necessarily helpful with a 166 - > technology like AI that uh has a global reach and we're not one 167 - > nation or one economic union can go and say, okay, this is how 168 - > we are going to do it and be completely unaffected by all the 169 - > other players in that field. 170 - > When you think back again to Davos, to the statements and 171 - > contributions that were shared from members from the European 172 - > Commission or other uh European institutions, anything that came 173 - > as a surprise to you or anything that you think will be 174 - > very relevant in the year ahead?

175 - > Kai Zenner: No, but uh well I was quite happy, uh, even though 176 - > um during Davos um there was a vote in the parliament on the 177 - > Mercosur um uh treaty or agreement, trade agreement, um 178 - > which um was perceived of course very badly also by the 179 - > Europeans in Davos. 180 - > But in general, I was um quite happy with the statements of um 181 - > European representatives there because you really saw that um 182 - > what I was just talking about, they understood that we are in a 183 - > very different geopolitical atmosphere and landscape, and 184 - > therefore you need to look for um new partners, maybe, but also 185 - > for strengthening um the relationships with existing um 186 - > countries or links that you have, but that are so far maybe 187 - > not so active.

188 - > So just go into one topic that the two of us know very well. 189 - > Um AI policy, um, the European AI Office so far struggled a 190 - > little bit when it comes to alignment with, for example, 191 - > South Korea, with Japan, with Singapore, and I think 192 - > especially those countries, we should now go basically all in 193 - > and uh try to align extremely closely. 194 - > And therefore, at least on when it comes to those high-level 195 - > statements in Davos, but also in at conferences before and 196 - > afterwards, um, our high-level policymakers, the prime 197 - > ministers, and so on are using or are making all the right 198 - > statements, which again I think is very positive.

199 - > Where I still would criticize uh the European Commission, but 200 - > also the member states, is when it comes really to practical 201 - > follow-ups, um they are often lacking this energy that the 202 - > official statements uh are indicating. 203 - > So there are not concrete steps, uh what to do. 204 - > And uh this brings me to a long-term criticism that um I 205 - > have, but also Axel Voss, my boss, has that an overall 206 - > strategy is missing for the digital field, but also for 207 - > other fields.

208 - > There are um going back to trade policy, um, even though 209 - > now there, I guess we have a clear strategy, we have 210 - > identified countries that we want to work uh closer with. 211 - > This strategy is not really shared with those colleagues 212 - > that are working in digital, that are working on 213 - > sustainability, that are working in law enforcement and so on. 214 - > So in this geopolitical world, we would finally need to 215 - > basically go out of the silos, work silos, and combine 216 - > everything.

217 - > And this unfortunately, so far, the European Union, but also 218 - > the individual member states uh are really lacking. 219 - > We are still in this kind of 19th-century policy approach 220 - > where you uh deal with, I don't know, with your car companies, 221 - > but only then with a very limited way, and you are not 222 - > combining it with other parts of the industrial environment, for 223 - > example. 224 - > Alexandra Ebert: So definitely uh not that helpful. 225 - > And also, as you pointed out, the member states, I think, 226 - > particularly the last few years, we've observed that they're 227 - > more like federalistic tendencies.

228 - > And particularly when we think about the digital space and the 229 - > scaling laws and everything that a flourishing ecosystem needs, 230 - > I think that's one of the areas that is currently underutilized 231 - > for the European competitiveness, to not have 13 232 - > uh AI factories distributed across the continent, but really 233 - > think about a united European approach. 234 - > Kai Zenner: Yeah, indeed. 235 - > And um, to become a little bit more positive uh when speaking 236 - > about the public sector, um I think in Germany, and Axel and I 237 - > have criticized the German ministries very often um for 238 - > their siloed approach, but um I think in Berlin many people do 239 - > share this frustration, and the new digital ministry is going in 240 - > exactly the um right direction.

241 - > Yes, there are some problems um because they are just building 242 - > it up, they need to figure out how this new player uh should 243 - > act in also uh among the other ministries, how they can 244 - > cooperate, what is really their role. 245 - > But even though it's early days, I would say here in 246 - > Brussels, you already see that Germany on digital policy has a 247 - > much more streamlined approach. 248 - > Finally, um, the German government could propose a joint 249 - > opinion before the digital omnibus.

250 - > Um and I would say credit to the digital ministry because 251 - > they managed to um bring both political parties of the 252 - > government together, but also um bring all the ideas from 253 - > different ministries together, and then send a joint position 254 - > to the Commission, and it had quite a lot of impact in the 255 - > Commission thinking and also in the final omnibus proposal of 256 - > the Commission. 257 - > So I think this example shows how powerful you could be if you 258 - > are breaking up the silos, and it's good to see that Germany 259 - > made this important step in the digital field.

260 - > Alexandra Ebert: Definitely. 261 - > And I think even though we hear doomsday reporting of okay, the 262 - > European Union is so far behind, we will never ever catch 263 - > up. 264 - > I don't share that sentiment. 265 - > Of course, there's massive work that needs to be done, but I 266 - > think there's still an opportunity, even though 267 - > urgency, of course, in certain areas is increasing, and also 268 - > the need to finally stand on our own legs and not only follow 269 - > narratives that are shared predominantly from tech, big 270 - > tech players in the United States and the strategies that 271 - > we see there and in China.

272 - > So uh still an optimist that the European Union will find its 273 - > European way, which maybe is a nice transition to a paper that 274 - > you co-authored uh last year, I believe, together with 275 - > the TUM think tank uh from the Technical University in Munich, 276 - > titled Uh, The European Way: A Blueprint for Reclaiming Our 277 - > Digital Future. 278 - > For those of our listeners who haven't read the paper, what are 279 - > the key recommendations that were shared to enable the 280 - > European Union and the European continent to do exactly this 281 - > reclaiming of our digital future?

282 - > Kai Zenner: Yeah. 283 - > Um, to give a little bit of um background, or let's say 284 - > historical background uh for this paper. 285 - > So, as I said, I'm now um over 10 years in Brussels and over 286 - > eight years um in the European Parliament, and uh Axe and I saw 287 - > a lot of positive sides, but also a lot of uh negative or 288 - > downsides when you are involved in in policy making, and we 289 - > share some frustration that certain flaws are not really 290 - > fixed, even though most people have identified the flaws, and 291 - > um countermeasures are available and also known.

292 - > And um, yeah, now going concretely to the paper, um I 293 - > had for a long time um an issue with the commission's use of the 294 - > label European way. 295 - > Uh, you know it for example, the European way uh for AI, um 296 - > AI made in Europe, the um EU AI policy, and so on. 297 - > And um in in keynotes, um Ursula von der Leyen, the 298 - > Commission president, or back then Vestager or Commissioner 299 - > Breton were always using this term, but there was no 300 - > substance.

301 - > No one really knew what is the third way to do digital or to do 302 - > AI. 303 - > What makes us in Europe special? 304 - > Why we are different uh in comparison to the US, to China, 305 - > to Israel, to India, to wherever. 306 - > Um, and in my opinion, by using this word, but by not being 307 - > able to explain it, you create a lot of confusion.

308 - > And of course, you you are not really making a good argument 309 - > also on an international stage why other countries should 310 - > cooperate with art, why our um digital services, technologies, 311 - > infrastructure is maybe superior in certain areas and so on. 312 - > And it also is connected to one point that I already made that 313 - > in particular digital field, um, there are not really long-term 314 - > plans. 315 - > We don't know where we want to go to um what um our ambitions 316 - > are, uh policy objectives, um, and so on.

317 - > So um, since I shared this frustration for so many years, 318 - > uh in 2024, so with the new election, um, I started to reach 319 - > out to certain allies that I um had in at think tanks, but also 320 - > um in some trade associations and so on and so on. 321 - > And um were telling those people, okay, look, um, in our 322 - > personal capacity as a kind of side project that we can do over 323 - > the weekends or in the nights, uh, let's just bring together 324 - > who needs sleep, who needs sleep.

325 - > Alexandra Ebert: Exactly. 326 - > Kai Zenner: Exactly. 327 - > But let's bring our frustration together and let's try to fill 328 - > those gaps. 329 - > And um, in the end, the whole paper had three key ideas or key 330 - > ambitions that I partially talked already about.

331 - > First of all, we finally wanted to fill the void, fill the gap. 332 - > What is meant by European way? 333 - > So we were coming up with six core principles. 334 - > I'm just naming now a few, things like interoperability, 335 - > things like trustworthiness, things like also sustainability, 336 - > and so on.

337 - > So points where you could rightfully claim that European 338 - > digital products, services, and so on are going in a bit of a 339 - > different direction compared, for example, with uh product 340 - > from let's say Microsoft and so on. 341 - > Maybe we are not as powerful, not as fast, not as whatever, 342 - > but we have other strengths. 343 - > For example, using less energy. 344 - > Um, those six principles, core principles, we kind of collected 345 - > based on our knowledge, experience, but also based on a 346 - > lot of policy documents.

347 - > Because the Commission and other institutions did a lot of 348 - > work in it, but also there in a siloed way. 349 - > There's, for example, the digital decade, which is giving 350 - > objectives for member states uh to transform their societies, 351 - > systems, and so on. 352 - > There's a declaration of digital rights and principles, 353 - > which is uh uh creating principles uh to protect um uh 354 - > the citizens, but again, it's all limited perspectives, it's 355 - > all certain areas only, and it's not this overall approach that 356 - > we were trying to do with the paper.

357 - > Then we um created a very strategic approach um of using 358 - > these new principles along the whole tech stack of the European 359 - > Union. 360 - > And as I said, um, based on our assumption that no one really 361 - > has an overall um policy strategy, we really tried in a 362 - > very high level way to go through the different layers. 363 - > And proposed a very strategic tactic of concentrating all of 364 - > our firepower basically in those areas where we could lead in 365 - > the future, in the digital field, where we are already 366 - > leading, and so on.

367 - > But for other areas and kind of taking a SWOT analysis approach 368 - > where we are not doing so good, or where we kind of lost 369 - > completely the connection in this kind of tech race, that we 370 - > are identifying international partners that are sharing our 371 - > values, or that are at least not hostile against us, and so on. 372 - > And the third point, and then I'm finally stopping of our very 373 - > broad table, was of course, all of that is not really feasible 374 - > in the current state, because all of us also shared strong 375 - > frustration when it comes to reforms.

376 - > We are talking uh about fostering digital skills. 377 - > Since I'm in Brussels, and long before there's basically almost 378 - > no step forward, we are talking about obstacles in the digital 379 - > uh single market in Europe. 380 - > Uh, you know, all the talk about um VCs not really helping 381 - > scale-ups and so on. 382 - > A lot of points that you find in the Draghi report and at 383 - > Letta report.

384 - > And we also put those together in a list of potential reforms, 385 - > but particular for the digital field and again connected to our 386 - > European way principles. 387 - > So stopping here, but this was our idea. 388 - > Alexandra Ebert: No, it was uh it was a really uh well-worth 389 - > the time read when I looked at the paper, because as you 390 - > pointed out, this holistic approach, this strategic 391 - > long-term vision is definitely something where I think the 392 - > European Commission could take a few uh additional steps towards 393 - > that.

394 - > And particularly also when we think about how to balance 395 - > long-term initiatives like working on digital sovereignty 396 - > in terms of uh compute infrastructure, in terms of 397 - > chips, of course, from a geopolitical perspective, this 398 - > is something the European Union should pursue. 399 - > But we know with the state of the art of development in this 400 - > highly technical, complex space where manufacturers are 401 - > operating at the edges of what is physically possible.

402 - > Uh, I think it's an illusion to say, okay, we're going to see 403 - > massive steps forwards over the next one, two, three, four, five 404 - > years. 405 - > So that's definitely a 15 years, 20 years endeavor to ever 406 - > get on par with uh leading edge providers. 407 - > And I think that needs to be balanced with things that 408 - > positively impact our economy, our society already in the next 409 - > one to five years. 410 - > And particularly there, I think what we're missing is building 411 - > up these flourishing ecosystems where scale-ups can grow, where 412 - > people have the resources, the data access, what they need to 413 - > actually have a meaningful adoption of AI.

414 - > And I think this role in the world is still up for grabs to 415 - > become a world champion of actually using AI to have a 416 - > benefit. 417 - > Yes, there are some uh leading edge models in certain areas, 418 - > but the latest frontier models and LLMs are just one category 419 - > of AI for a functioning economy and a benefit from it. 420 - > You don't just need the, I don't know, Boeing 747s for 421 - > everything. 422 - > You need the small truck, you need the bike, you need the baby 423 - > stroller, you need the car.

424 - > And making this adoption happen for particularly the European 425 - > medium-sized enterprises, small enterprises, which are our 426 - > economy's backbone, I think would be a fantastic focus area 427 - > for the entire European Union to make sure that we are already 428 - > seeing positive impact in the next handful of years ahead, and 429 - > not only with this long-term, big, big, big, geopolitically 430 - > important but not very uh easily um realizable plans in terms of 431 - > tax sovereignty.

432 - > Kai Zenner: If I can compliment uh what you are saying, um 433 - > please when we are now um going really back to the AI uh core 434 - > field, it's exactly what you're saying, and again bringing back 435 - > a SWOT analysis. 436 - > Instead now of um funding certain moonshot uh projects, 437 - > but all with rather small sums like we did in the past in 438 - > Horizon, for example, and so on. 439 - > Um why not really looking or using also foresight a little 440 - > bit uh to see where we could lead?

441 - > Maybe it's not worth anymore to put all of our money um in um 442 - > large language models because, as you said, it's just one field 443 - > of many. 444 - > Maybe there are other areas where um it's much better for 445 - > us. 446 - > And again, also we um never in our uh paper argued for 447 - > protectionism for outer key of the European Union. 448 - > So we want still to work with our US partners, with other 449 - > partners all over the world, and again complement where we are 450 - > strong with areas where they are strong.

451 - > And if you are using the SWOT analysis um for your, for 452 - > example, broader AI policy, I think we could really make 453 - > finally a step forward and again would not waste so much 454 - > political capital but also financial capital by spreading 455 - > everywhere. 456 - > And the other thing that is also very well connected to what 457 - > you're just saying, and again, there is a very positive step, 458 - > but we are definitely not there yet. 459 - > You were talking about those hubs of excellence or clusters 460 - > of excellence and so on.

461 - > Um, the European Union is trying to go in this direction 462 - > for quite some time. 463 - > As you know, there are those EU HPCs, supercomputers. 464 - > We have the uh European digital innovation hubs that are all 465 - > over Europe, uh, but all over Europe is exactly the problem 466 - > because uh the digital innovation hubs are not 467 - > connected um at all to other existing clusters. 468 - > Uh let it be now the AI factories or the gigafactories.

469 - > Now soon there will be data labs, there will be regulatory 470 - > sandboxes. 471 - > Again, they are also the um digital innovation hubs, the 472 - > supercomputers, all of those things are often managed by 473 - > different ministries, by different parts of the European 474 - > Commission. 475 - > No connection between them. 476 - > And this is such a waste of um um human capital, of um 477 - > financial means, and you will never manage what we should 478 - > build.

479 - > We should build, of course, a decentralized system of 480 - > excellence in different digital fields, but those fields need to 481 - > be connected. 482 - > So, just to give you an example, in Hessen, in the 483 - > middle of Germany, around Frankfurt, the state, they um 484 - > are very strong with the data centers because there are the 485 - > data streams coming to uh Europe and Germany. 486 - > Um, of course, they are a financial center. 487 - > So, probably if you want to do something in digital, those two 488 - > areas makes a lot of sense.

489 - > If you now want to do machinery stuff or automotive industry, 490 - > well, why waste energy there? 491 - > Why not do it in Baden-Württemberg and Bavaria? 492 - > But for all of that, you need again an overarching plan, and 493 - > you need to make sure that all those uh hubs are in the end 494 - > coordinating and connected, and not what unfortunately seems now 495 - > to happen is that the AI factory um in Baden-Württemberg 496 - > uh will go in one town that is 200 kilometers away from um the 497 - > uh regulatory sandbox and so on.

498 - > And then, of course, you don't you are not able to leverage uh 499 - > synergies. 500 - > Uh you are not able to provide a startup with one place where 501 - > they can find everything. 502 - > For example, the data protection authority that is 503 - > helping them as a service provider to become GDPR 504 - > compliant. 505 - > At the same time, they could use the SME trainer from a 506 - > digital innovation hub that is helping them to boost um and to 507 - > scale up their product.

508 - > They don't have uh other authorities that are helping 509 - > them to become compliant with the AI Act um to help them to um 510 - > uh to find uh funders and so on. 511 - > And yeah, I I really hope that we have all those things often 512 - > already existing, but again, we need to bring them together. 513 - > Bring it together. 514 - > And I really hope that this is happening in the future.

515 - > Alexandra Ebert: Let's let's hope that 2026 is uh is the year 516 - > where that's going to change. 517 - > I mean, you shared so much to reconnect to, particularly to 518 - > the second point that you made about also the waste of 519 - > resources or inefficient use of resources, the financial means. 520 - > That's also a critique point I've been sharing with media and 521 - > and uh organizations I'm advising over the last year. 522 - > That I have the impression that particularly on the European 523 - > stage, but also in the UK quite similarly, there is this hyper 524 - > focus on building European frontier models and LLMs, 525 - > granting ever more computer resources there, even though, 526 - > from a scientific point of view, and also from what we're seeing 527 - > in other economies, the verdict is still out, and many experts 528 - > are even doubting that the way uh the compute intense and 529 - > data-intense way that AI frontier models are developed 530 - > and trained today is really the last uh way of uh AI 531 - > architecture that will have the most societal and economical 532 - > impact.

533 - > Yet it's an extremely costly way. 534 - > And therefore, I think, particularly as a public sector, 535 - > government, regulator, political leader, you have this 536 - > responsibility towards your citizen to also assess the 537 - > opportunity costs of should I really spend billions in compute 538 - > infrastructure and this attempt to have an LLM that's trained 539 - > on European values, when we're not yet seeing strong indicators 540 - > that LLMs are really having economical upside or societal 541 - > upside.

542 - > So I think there again is a strong argument to think more in 543 - > terms of building up an ecosystem where we can have this 544 - > whole set of vehicles being developed in a meaningful 545 - > capacity for our small and medium uh businesses where we 546 - > can have uh scale-up support, or when we even think about the 547 - > economic um hotspots in the United States, in uh China, 548 - > where you have all of these providers and mid-sized 549 - > enterprises, particularly in China, that are so quick to 550 - > develop a prototype versus talking with startups in the 551 - > European Union, it is extremely costly, it takes ages until you 552 - > have the first uh prototype.

553 - > So I think thinking more on an ecosystem layer, centralizing 554 - > our endeavors there and our approaches is something that 555 - > will be needed to bring the European Union on a competitive, 556 - > flourishing ecosystem stage. 557 - > Kai Zenner: Indeed. 558 - > And especially now that you are talking about foundation 559 - > models, um, I think it it's probably still worthy to invest, 560 - > to have some companies from Europe that are leading there or 561 - > that are active.

562 - > But again, is it really useful to replicate the leading Chinese 563 - > models and the leading US models? 564 - > Um I would say no, and therefore I like what Mistral uh 565 - > but also Silo AI, which is now part of A M D, um, are doing um 566 - > Black Forest Labs in Baden-Württemberg also. 567 - > And one of the first was actually Falcon from the 568 - > Emirates from Abu Dhabi, um, which are going in a different 569 - > direction. 570 - > They don't want um by any means the to create the best, the 571 - > fastest, the most powerful um um foundation model, but they are 572 - > saying um we want, and you mentioned it also slightly, we 573 - > want to have uh a foundation model that is, for example, um 574 - > rather fluent in the Romanian language, in the Romanian 575 - > culture, has a lot of um insight in certain special ethical um 576 - > values, um approaches, and so on.

577 - > Uh the Falcon model I found really interesting when I heard 578 - > it for the first time that they have direct access to a lot of 579 - > libraries in the um Emirates, but also in the whole Arabic uh 580 - > um um area, um, and similar things um those foundation model 581 - > providers uh from Europe are now trying to do, which makes a 582 - > lot of sense because, as you know, over 65% of the content in 583 - > the internet um are in English. 584 - > We in Europe have a lot of diversity when it comes to 585 - > languages, um, so this is a perfect niche for companies um 586 - > to strive into because it's probably not something that US 587 - > models will ever really focus on.

588 - > So there is a gap that we could fill where we could still earn 589 - > money, but also provide our society, um our state something 590 - > that is really worthy and actually also worthy to protect 591 - > in order to not only have now English content to dominate 592 - > everything. 593 - > So, this I think this example that you gave is really, really 594 - > great to show yes, we should first not only replicate, we 595 - > should also not run behind and focus everything on one 596 - > technology, but even in one technology where others are 597 - > dominating, there are still niches which we could fill.

598 - > Alexandra Ebert: Absolutely. 599 - > And particularly when we think about large language models also 600 - > having this big advantage of being a natural language user 601 - > interface that by the state that we currently have is something 602 - > that you can plug into other AI systems to benefit from the 603 - > skills without having to train it uh from scratch again. 604 - > When we build on the example, when I build on the example that 605 - > you just share about doing these models language 606 - > specifically and maybe more in tune with the diversity we have 607 - > in Europe, I think another really, really interesting and 608 - > not yet uh economically utilized area is the proprietary 609 - > business data.

610 - > We see it with OpenAI, we see it with Anthropic and many of 611 - > those players who are now focusing much more on European 612 - > businesses with their forward-deployed engineers 613 - > helping them find privacy-compliant ways to get 614 - > all of the proprietary data. 615 - > But if you look at that at scale, that's also an 616 - > opportunity for those US players to gain an enormous data 617 - > advantage. 618 - > And I think here it would also be important for the European 619 - > Union to think about how to get this type of data accessible to 620 - > European players because there's a whole different area of value 621 - > to be realized in an economical perspective when you get access 622 - > to proprietary business data, because there's only so much 623 - > that a generically trained ChatGPT can tell the 624 - > medium-sized enterprise about their specific customers.

625 - > So I think that's actually a new frontier that uh is not 626 - > necessarily discussed enough. 627 - > How we can keep this data advantage there and building 628 - > upon that, also, how can we open up data assets, particularly 629 - > for the societal impact that AI could have in the health space, 630 - > in the financial inclusion space, in the climate space, 631 - > because so many of the solutions that would be tremendously 632 - > beneficial on the public sector side of things and on the public 633 - > goods side of things are not necessarily something that 634 - > builds revenue for specific organizations and therefore 635 - > might not be pursued by those who have access to this data to 636 - > develop it because there's no business motive behind it.

637 - > So this is why I've been advising and advocating so much 638 - > for also this uh open data layer and societal layer of utilizing 639 - > data and making sure the resource is there for 640 - > researchers, NGOs, and public sector representatives to grab 641 - > and make something that really moves the needle on these 642 - > important issues. 643 - > Kai Zenner: No, it's a it's a great point. 644 - > And also here there's good news and bad news. 645 - > I start with the good news.

646 - > Uh as you is as you are aware, um, the European Union has 647 - > identified um this issue a long time ago. 648 - > Um, in the last legislative term, they proposed the uh data 649 - > strategy in 2020. 650 - > Um that was together with um the Data Act and the Data 651 - > Governance Act, trying to create European data spaces that are 652 - > all interoperable and so on. 653 - > Yeah, I do believe that the concept, in a theoretical 654 - > perspective, is really good.

655 - > In practical terms, as you know, especially industry, 656 - > didn't really welcome it. 657 - > Uh, there was a lot of pushback, and so the pushback 658 - > mainly focused on IP, on trade secrets. 659 - > Companies in Europe didn't really want to um to share 660 - > insights, uh, certain competitive advantages with 661 - > others, also between sectors. 662 - > Um, it didn't really work well then, of course, in Europe, um, 663 - > even more than in other areas.

664 - > We have this issue with uh privacy uh standards, with 665 - > copyright uh protection, and I would say especially those three 666 - > points really prevented us to, as you said, leverage this 667 - > treasure that we have in Europe, also when it comes to 668 - > industrial data and all other data sets. 669 - > Uh the commission also tried to learn from that, and um, just 670 - > recently in November last year, um, proposed a new data union 671 - > strategy that wants to solve certain things, and there is now 672 - > the data.

673 - > No, I know, and I share your frustration. 674 - > And there is now the digital omnibus that is trying to um 675 - > merge basically Data Act and uh Data Governance Act. 676 - > Um, so at least they identified the problem, they acknowledge 677 - > the problem, they are trying to solve it. 678 - > But as you said, and this brings me to the bad news, I 679 - > share your frustration.

680 - > I'm also not really convinced that those proposals will help 681 - > to solve the three issues, main issues that I mentioned. 682 - > And in particular, if you talk with industry, they completely 683 - > share our assumption and also say, Well, uh, no, I still don't 684 - > want to uh share my data sets and so on and so on. 685 - > So, there I think we really need we need more engagement 686 - > from the public sector with industry. 687 - > Of course, we need to be also, we need to demand more from 688 - > industry to rethink things, but I believe the public sector also 689 - > they cannot dictate it if they're trying and again and 690 - > again.

691 - > Uh they need to find a joint understanding uh with industry 692 - > players, and I think going back to our geopolitical discussion 693 - > at the beginning, it's now or never. 694 - > No, I think if if we cannot make find this agreement now, 695 - > yeah, then we lost. 696 - > Alexandra Ebert: I mean, I agree with many points that you said, 697 - > but I disagree, or I think it's an even more like fundamental 698 - > issue that's at play here. 699 - > So, yes, the data spaces on a very high-level perspective are 700 - > a good idea, but when we look more closely into how they're 701 - > set up currently, and also what the Data Act and the Data 702 - > Governance Act did in terms of uh framework, uh incentive 703 - > structures which are predominantly lacking, the data 704 - > intermediary role and uh and all of that.

705 - > I think this just is a sign of the immaturity when it comes to 706 - > using data as a strategic asset. 707 - > And that's not necessarily a complaint or something that's 708 - > unique to the European Union. 709 - > I'm not aware of any country in the world that really has a 710 - > holistic framework about how data can be used as a strategic 711 - > asset, both in the economy but also beyond that for societies 712 - > at large. 713 - > I mean, yes, China made it a uh a production factor factor, I 714 - > believe, like four years, five years ago, but still, this 715 - > framework on how could that work within a nation, within an 716 - > economic union globally, is not yet developed.

717 - > And I think we need to bring together forward thinkers, 718 - > economists, industry representatives, public sector 719 - > uh experts who can really come up with a new way that has this 720 - > incentive mechanisms that discusses how the value chain 721 - > could look like that. 722 - > Even maybe toys with uh provocative ideas like data as a 723 - > tax or something like that to open up the resource. 724 - > Because we must face it, data is an asset, it's an economic 725 - > asset, but it's fundamentally different than natural resources 726 - > or money where we have very mature um uh Ecosystems and 727 - > means to treat them as a strategic asset.

728 - > Because one simple example, even though a natural resource 729 - > like gas is something that either you can use or I can use, 730 - > the uh gas can be in your car's tank or in my gas tank. 731 - > We can't drive with the same uh liter of gas at the same time. 732 - > Data has this non-rivalry element where you can use it 733 - > economically, you can share it with a partner organization, you 734 - > can give and give it up to a public sector organization or 735 - > researcher to do good with it.

736 - > And I think this needs new thinking. 737 - > And I'm personally not aware that there's a really thorough 738 - > debate happening how this future of the data economy could look 739 - > like in the European Union or anywhere else in the world. 740 - > Kai Zenner: I would claim, and it's a place you also know very 741 - > well, that probably in the Western Hemisphere or in this 742 - > Western Alliance, uh the only place where this is being done 743 - > is the OECD in Paris.

744 - > They are thinking a little bit about new ideas. 745 - > Alexandra Ebert: I've actually brought that to their attention, 746 - > and they were quite receptive to say, okay, you're right, this 747 - > specific aspect uh maybe needs more attention. 748 - > So let's see, maybe there will be an additional expert group 749 - > specifically on that. 750 - > Kai Zenner: And I know it also from the AI field.

751 - > Um, they were quite successful in in shaping the the whole 752 - > European AI Act, at least when it comes to the principles and 753 - > so on. 754 - > But yeah, let's say, besides um or beyond um OECD, I completely 755 - > agree that there is no platform that is really doing this kind 756 - > of foresight and um agenda setting. 757 - > Uh and I would therefore also conclude that in Germany is a 758 - > democratic system, but also here in the in the whole uh European 759 - > uh Union is really suffering under this non-existing of free 760 - > thinking in foresight, in in um agenda setting, um coming up 761 - > with new ideas.

762 - > We saw it in the Draghi report and um in the Letta report, 763 - > there were again a lot of good um a lot of good assessments, 764 - > evaluations of status quo, and both of them made some new 765 - > ideas, but how did policymakers react? 766 - > Basically, by just continuing to do what they are doing, yeah. 767 - > Sometimes they just relabeled old concepts, um, so EU HPCs 768 - > became AI factories, for example, and so on. 769 - > And indeed, I agree there with you.

770 - > What is missing is a really starting from scratch um 771 - > movement, basically, or um group of people uh who are locking 772 - > themselves in a room, and for example, in one area where Axe 773 - > and I are super frustrated, data protection are saying, Okay, 774 - > look, it's not all bad. 775 - > We can be proud of our high privacy standards and so on, but 776 - > certain things, uh what Axe is always saying worked in the 80s, 777 - > especially before AI. 778 - > But nowadays, is it for example the principle uh of data 779 - > minimization?

780 - > Is it really useful to have it nowadays? 781 - > Or if we don't want to uh uh scrap it away completely, maybe 782 - > we should reinterpret it. 783 - > But no one is doing it, and we could now go to a lot of other 784 - > areas, even to the AI area where I completely agree with you. 785 - > We should be more radical.

786 - > Axel is saying it for now very uh many years. 787 - > We need to be radical, we need to a bit like with Foucault, the 788 - > French philosopher, we need to reconstruct our score our used 789 - > concepts and really rethink them and um uh assess um are they 790 - > really beneficial anymore? 791 - > Do we need complete new approaches? 792 - > And this brings me back to the example that uh you made.

793 - > Also, there is what we are doing working. 794 - > We both would say no. 795 - > So, not our options, yeah. 796 - > And then we have maybe a list of five options, and then maybe 797 - > also do try and error.

798 - > Try out. 799 - > Exactly. 800 - > Exactly. 801 - > If something is not working, or best would be to use two or 802 - > three strategies in parallel, and then survival of the fittest 803 - > is one absolutely one of them.

804 - > Alexandra Ebert: That's that's such an important point you're 805 - > making because, from a technological perspective, we 806 - > are already way more mature in terms of what is possible when 807 - > we think of privacy, when we think of data minimization, when 808 - > we think about all of these things that could help us to a 809 - > flourishing data ecosystem. 810 - > It's way more this question of paradigm shift and change 811 - > management and the human factor that needs to be addressed 812 - > because, on the one hand, most uh representatives, political 813 - > leaders don't even grasp what the challenge is here and what 814 - > is at stake, or it's uh novelty and not being familiar with the 815 - > different uh technologies that maybe keep them in the status 816 - > quo.

817 - > So I think technologically we have more than enough to try out 818 - > and see out how we can achieve a new European way and 819 - > competitiveness that isn't done anywhere in the world, but we 820 - > need to be brave enough to make this calculated risks, to make 821 - > these steps, and to experiment uh with something. 822 - > So uh definitely uh a big mission ahead to change the 823 - > status quo. 824 - > Because we have to come to an end. 825 - > Uh, maybe let's close with uh two big questions.

826 - > We are just still at the beginning of the year 2026. 827 - > So when you think ahead, what the areas you and Axel are 828 - > working on that are most exciting to you and that you 829 - > think will either have the biggest upward benefit of 830 - > propelling the European uh Union forward in all of the topics we 831 - > talked about, or maybe that you're also concerned about 832 - > because depending on how they will play out, they could also 833 - > uh push us backwards and be an obstacle.

834 - > What would that be? 835 - > What is uh kind of top of mind for you and most important in 836 - > 2026 in your digital policy? 837 - > Kai Zenner: It's a great question, and I'm so happy that 838 - > you uh give me the chance to talk about two points that I'm 839 - > really passionate about. 840 - > And um let me just think how I structure it.

841 - > Maybe I start with the negative one in order to give us a 842 - > positive ending ending, exactly. 843 - > Like you're thinking. 844 - > So the negative one, um, and I kind of indicated it in in 845 - > several parts of um our recording. 846 - > Um Axel and I are extremely frustrated, um, especially here 847 - > in Brussels, how yeah, we are calling it the policy machine or 848 - > the policy cycle is um accelerated.

849 - > We are now for so long in crisis mode, and we just talked 850 - > about it. 851 - > Things like foresight, things like evidence-based policy 852 - > making do not really play a role um so much. 853 - > It's all about um finding quick fixes and so on. 854 - > And we are very, very afraid that even though the Draghi 855 - > reports, the Letta report, many NGOs, many think tanks, many 856 - > trade associations are all indicating the same things, are 857 - > all showing at least in a similar direction, the EU policy 858 - > machine is not really able to respond because they are kind of 859 - > blocking themselves.

860 - > Um the too many people um are depending on maintaining the 861 - > status quo in order to keep their power, their competences, 862 - > they are not open for new ideas, they lack the courage, how we 863 - > say here in Brussels, um to really do bold steps. 864 - > Um and yeah, I think we need really bold policy leaders, um, 865 - > like Kohl and Mitterrand in the 90s, uh um like Adenauer, um uh 866 - > like Helmut Schmidt now to name a few Germans that um took a lot 867 - > of risks, personal risks, um, also risks in their career.

868 - > Um, but those risks are important to to make big steps 869 - > forward, like also the euro and so on. 870 - > Um, and I think, and Axel shares that we need things like 871 - > the euro nowadays in the digital field, in other fields. 872 - > And again, we don't we don't see the momentum there. 873 - > And um maybe it needs to come from the outside at the moment.

874 - > And therefore, again, we found the Drari report, the letter 875 - > report, but also other initiatives extremely helpful 876 - > because maybe with those external uh pushes um we will 877 - > build up momentum for someone to find the courage to step 878 - > forward and not think anymore so small scale only for his or her 879 - > own member state or for certain advantages over other um areas 880 - > of the world. 881 - > No, we need now big thinkers, big reformers, and so on.

882 - > So this was the more negative point. 883 - > Um the positive point um for this year, for next year, and so 884 - > on is really the mentioned strategic thinking that I um 885 - > talked about at the very beginning, that is finally 886 - > coming back to Brussels, finally coming back uh to policy. 887 - > And as someone who again was there from the start when it 888 - > comes to the Eurostack movement, but Axe and I are working on 889 - > tech um resilience um and strategic uh policy making for 890 - > all the time we work together in Brussels.

891 - > It's so great to see that after being in a niche for those 892 - > eight years we work together, now we are on the main stage. 893 - > Front and center, yes. 894 - > Exactly. 895 - > There are hundreds.

896 - > Davos was all full about uh geopol uh politics, about 897 - > strategic approaches that Europe needs to take, and so on. 898 - > Um, after I bashed them here and there now in our recording, 899 - > the commission actually there also changed the course. 900 - > Things like buy European about European preference. 901 - > Again, we don't want to become protectionist, but maybe in 902 - > certain areas it makes sense.

903 - > The commission very likely will make um a proposal to change 904 - > our public procurement directives in this area. 905 - > Also, the AI Cloud and Development Act um is go uh 906 - > cloud and AI Development Act is going in this direction, and ACE 907 - > and I and our network are very supportive of it, and um we are 908 - > very relieved that our pushes in this area finally uh show some 909 - > merit success. 910 - > Exactly. 911 - > And uh yeah, this basically makes my outlook rather positive 912 - > because I think it's just a starting point, or hopefully, 913 - > just a starting point, and it will get even more.

914 - > And this is a very, very positive um development. 915 - > Maybe that leads also to us overcoming my negative point the 916 - > lack of political leadership or reforms, because I think the 917 - > more we are going in this very strategic uh direction, the more 918 - > leverage or the more um means our political leaders also have 919 - > to bring up and make those decisive steps. 920 - > Alexandra Ebert: Well, very, very happy to hear your 921 - > optimism, and as we discussed it, I think we both both share 922 - > that, so it's definitely going to be an exciting year.

923 - > And just one thought I wanted to share uh you mentioning the 924 - > procurement, I think that's also something so interesting to 925 - > look at because when one looks back into the history books of 926 - > the secret source of the Silicon Valleys and uh Israel's 927 - > innovations, strengths of the world, I think uh in many of 928 - > those areas, it was actually this industrial military complex 929 - > that propelled the ecosystems forward. 930 - > And in a public sector setting, in a government setting, 931 - > historically speaking, it was almost always the military 932 - > component that would was actually the funder of the most 933 - > frontier innovation because uh of the way they approach this 934 - > because of the unique needs.

935 - > With the procurement we've had in in Europe for the past 936 - > decades, uh, this need to say, okay, please show me three 937 - > different or how many ever different players with a 938 - > comparable offer, per definition means we're not talking about 939 - > frontier innovation and something that's so unique that 940 - > there is per definition no competitors. 941 - > So also thinking about that and how the spending power that we 942 - > have on the public sector can be utilized to build up an 943 - > ecosystem and maybe not only go to the big players who are 944 - > always getting it, but really making smart decisions that 945 - > could help us replicate some of these ecosystem building um 946 - > elements we saw working in other areas of the world is something 947 - > that I think is an underutilized area for the 948 - > European innovation ecosystem.

949 - > Kai Zenner: Definitely. 950 - > And I know we are uh very close to the end, but just one 951 - > example which is going directly to what you are saying. 952 - > I just recently heard, I think it was actually in Davos, um, a 953 - > CEO told me from a startup um that public or how important 954 - > public procurement is. 955 - > Uh, she told me um that in the end uh they got a small contract 956 - > from I think it was France of 20 million uh euros.

957 - > Um so for them it was really a small scale. 958 - > Um but by having this contract, they were kind of whitelisted 959 - > in the eyes of funders of other companies who were saying, Well, 960 - > if the French government is trusting this company so much to 961 - > give them this contract or the standard, then we can also take 962 - > it. 963 - > And basically it um triggered investments of so 20 million 964 - > triggered investments, I think to close 1 billion in the end, 965 - > or really an astonishing um thing, of course, not uh the 966 - > next year, but over the years.

967 - > And this shows that with not a lot of public investments, we 968 - > could in the end create European champions, or at least we could 969 - > um strengthen our economy, you know. 970 - > And yeah, indeed. 971 - > So, what you said is completely right, in my opinion. 972 - > Alexandra Ebert: Yeah, well, let's see where we will get uh 973 - > towards the the year ahead and if we will already see first 974 - > steps in this uh hopeful uh and positive direction for the 975 - > European Union.

976 - > Well, Kai, thank you so much. 977 - > It's always a pleasure to talk with you and I really enjoy the 978 - > depth of your thoughts and uh our exchanges. 979 - > So thank you so much for coming on the podcast. 980 - > And maybe already at the end of the year we have enough 981 - > material.

982 - > I'm pretty confident we would have to make a second episode 983 - > and see where we landed. 984 - > Kai Zenner: Thank you very much. 985 - > And yeah, let's do it. 986 - > Alexandra Ebert: Awesome, awesome.

987 - > Kai Zenner: See you soon. 988 - > Alexandra Ebert: What a fun conversation. 989 - > I hope you enjoyed it as much as I did. 990 - > After knowing Kai for, I believe, five years, so pretty 991 - > much uh since I started this podcast, it was phenomenal to 992 - > finally invite him on the show and not only have our 993 - > conversations in some speaker's backstage area when we meet at a 994 - > random conference and having the time and breath that his 995 - > former deluxe was definitely something I cherished.

996 - > In fact, I actually was in for a treat that day and also 997 - > interviewed him for an Austrian legal publication called ailex. 998 - > So, for the German listeners of this podcast, if you're curious 999 - > about a deep dive with Kai, particularly when it comes to 1000 - > the digital omnibus and the AI omnibus, do check out the March 1001 - > issue of that magazine. 1002 - > It's well worth your time. 1003 - > It's a really phenomenal read.

1004 - > With that said, thank you for tuning in today. 1005 - > Thank you for sticking around until this moment of the 1006 - > episode, and see you again or hear you again in two weeks' 1007 - > time.